Testing the Duke Population Health Profile (Duke-PH) in a Sample of Community Health Center Patients

Abstract

Introduction: Our purpose was to develop and test a brief, self-report, and scorable survey instrument for measuring population health profiles from the individual respondent's perspective. We defined population health as the state of physical, mental, and social well-being of a group of individuals, including determinants of their well-being.

Materials and Methods: Respondents were adult patients in a community health center. Instrument items, an overall scale, and two subscales were developed and evaluated. Reliability was tested by Cronbach's alphas and test-retest correlations; construct validity was tested by correlations between scores and economic and clinical factors; criterion validity was tested by regression analyses for prediction of morbidity and health care utilization by baseline scores; and feasibility was tested by length of administration time.

Results: This was a 2-years prospective study of 450 patients, mostly black non-Hispanics (54%) and Hispanics (29%), many with no health insurance (45%), and poor enough to meet the federal poverty level (73%). The Duke Population Health Profile (Duke-PH) was developed with a 14-item PH scale for overall population health profile and two 7-item subscales, one for social determinants and the other for health determinants. Validity of item selection was indicated by item convergent and item discriminant correlations. Scale and subscale reliability were supported for internal consistency by Cronbach's alphas of 0.63–0.73, and for temporal stability by test-retest correlations of 0.65–0.78. Support for construct validity was shown by the more favorable baseline subscale and scale mean scores for patients able to buy private insurance than for patients unable to afford it. Criterion validity was supported by regression analyses showing that baseline scale and subscale scores predicted both baseline morbidity and 6-months utilization. Feasibility was shown by the mean self-administration time of 3.9 min and mean interviewer-administration time of 5.8 min.

Discussion: The strength of this study is support for Duke-PH reliability, validity, and feasibility in a community health center patient population. The new instrument is unique because it measures both social and health determinants of population health from the perspective of individuals in the population.